I was reviewing this presentation at the IETF 87 by Laurent Vanbever (postdoc research from Princeton) and does a great job of pointing out some of defects and failures of the BGP protocol:
BGP reconfigurations can create:
signaling anomalies ( routing oscillations )
dissemination anomalies ( black holes )
forwarding anomalies ( forwarding loops & traffic shifts)
or any combination of those
Noteable point and proofs that “Best practices do not work” and “Tuning eBGP policies can create huge traffic shifts”.
Why is BGP reconfiguration so complex ? Local reconfiguration can have global impact in an unpredictable manner
I point this out because every time I criticise routing protocols like BGP or MPLS or OSPF, someone wants to argue that they are great because they work. And they must be good because that’s what the everyone uses. Blah, blah, blah …..
That’s a path to stagnation and failure. Keeping something “just because it works” means we should still be riding horses instead of driving electric cars. My view on BGP is best described by this saying from Thomas Edison:
“I have not failed. I’ve just found 10,000 ways that won’t work.”
It’s true that replacing BGP seems to be an impossible task but lets be honest and realise that BGP is 20 year old protocol that has serious problems and limitations that are not yet solved. We just hope that it keeps working and pretend that the technology foundations of networking are perfect.
Fortunately, Laurent proposes some changes to BGP that I suspect no one will want to implement. Correct me if am wrong in the comments.